Earnings questions have been asked in South Africa’s national surveys annually since 1994. A
key question for labour economists has been to track and explain the evolution of earnings over
this post-apartheid period. Unfortunately, however, the measurement instrument has changed
in ways that make it tricky to simply take the raw figures and compare them even if one restricts
the attention to the October Household Surveys and the various Labour Force Surveys. In this
paper we analyse some of the changes and indicate where corrections are needed. We implement
many of these changes in the second release of PALMS, the Post-Apartheid Labour Market Series
(Kerr, Lam and Wittenberg 2013).
The structure of the paper is as follows. In section 2 we review studies done on the earnings
variables in the national surveys from Statistics South Africa, particularly those that comment
on the comparability of the variables over time. In Section 3 we pay attention in more detail
to the evolution of the measurement instrument. We then turn to an analysis of the actual
responses in section 4 with a view to pinpointing where the underlying measurements may have
changed. The following sections deal with ways of handling bracket information and missing
data respectively . In section 8 we look at the impact of these data quality adjustments on the
estimation of average real earnings over time.